Suomi satellite brings to light a unique frontier ofnighttime environmental sensing capabilitiesSteven D. Millera,1, Stephen P. Millsb, Christopher D. Elvidgec, Daniel T. Lindseyd, Thomas F. Leee,and Jeffrey D. Hawkinse

aCooperative Institute for Research in the Atmosphere, Colorado State University, Fort Collins, CO 80523; bNorthrop Grumman Aerospace Systems, RedondoBeach, CA 90278; cNational Geophysical Data Center, National Oceanic and Atmospheric Administration, Boulder, CO 80305; dRegional and MesoscaleMeteorology Branch, Center for Satellite Applications and Research, National Environmental Satellite, Data, and Information Service, National Oceanicand Atmospheric Administration, Fort Collins, CO 80523; and eSatellite Meteorological Applications Section, Marine Meteorology Division, Naval ResearchLaboratory, Monterey, CA 93943

Edited by Michael Mishchenko, National Aeronautics and Space Administration Goddard Institute, New York, NY, and accepted by the Editorial BoardAugust 10, 2012 (received for review April 25, 2012)

Most environmental satellite radiometers use solar reflectanceinformation when it is available during the day but must resort atnight to emission signals from infrared bands, which offer poorsensitivity to low-level clouds and surface features. A few sensorscan take advantage of moonlight, but the inconsistent availabilityof the lunar source limits measurement utility. Here we show thatthe Day/Night Band (DNB) low-light visible sensor on the recentlylaunched Suomi National Polar-orbiting Partnership (NPP) satellitehas the unique ability to image cloud and surface features by way ofreflected airglow, starlight, and zodiacal light illumination. Examplescollected during new moon reveal not only meteorological andsurface features, but also the direct emission of airglow structures inthe mesosphere, including expansive regions of diffuse glow andwave patterns forced by tropospheric convection. The ability toleverage diffuse illumination sources for nocturnal environmentalsensing applications extends the advantages of visible-light in-formation to moonless nights.

airglow/nightglow | nocturnal remote sensing

The sky on a dark, moonless night is, in fact, immersed withina sea of visible-spectrum light that the dark-adjusted human eye

can only begin to discern. The primary sources are the polar aurora,airglow, integrated starlight (including the Milky Way), andzodiacal light (1–3). Auroras, although a relatively strong source,are ephemeral and confined to high latitudes. The other sourcesproduce a complex global distribution of nighttime diffuse skybrightness that varies considerably across space, time, and spec-trum. At visible and near infrared wavelengths (e.g., 0.4–1.1 μm),the combined illumination from these sources yields down-wellingradiances at Earth’s surface in the range ∼10−11 to 10−9 W·cm−2·sr−1·μm−1 (3), or approximately 1 billion times fainter than sunlight.Low-light imaging capabilities have existed on the Operational

Linescan System (OLS) on the Defense Meteorological SatelliteProgram (DMSP) constellation since the late 1960s. The OLS wasdesigned to amplify visible light and detect clouds under twilight andmoonlight (e.g., signals down to ∼10−8 W·cm−2·sr−1·μm−1) illumi-nation conditions (4) but soon revealed many additional capabilitiesbased on signals from both natural and anthropogenic sources (5–8).The Suomi National Polar-orbiting Partnership (NPP) satellite

(http://npp.gsfc.nasa.gov) was launched on October 28, 2011, andplaced into an 834-km altitude sun-synchronous orbit with localequatorial crossing times of ∼1:30 PM and 1:30 AM. Named inhonor of Verner E. Suomi, considered widely as the “father ofsatellite meteorology,” NPP provides risk reduction for the JointPolar Satellite System (JPSS) series of National Oceanic andAtmospheric Administration (NOAA) operational meteorolog-ical satellites and continuity to the National Aeronautics andSpace Administration (NASA) Earth Observing System (EOS)Terra and Aqua climate research satellites (9).Suomi NPP carries the Visible/Infrared Imager/Radiometer

Suite (VIIRS), an optical spectrum (22 bands spanning ∼0.4–13

μm) sensor providing imagery at high spatial resolution (0.375–1.6 km, band dependent) across a 3,000-km-wide swath. Includedon VIIRS is a next-generation low-light sensor, the Day/NightBand (DNB). The DNB has a measured spectral response of505–890 nm (full width at half maximum, with nominal band-center wavelength of 705 nm) and features several advances tothe heritage OLS, including full calibration and improved spatial(0.74 km vs. ∼3 km) and radiometric (14-bit vs. 6-bit) resolutions(10). Three stages of gain allow the DNB to span the dynamicrange of radiances encountered during the daytime, twilight, andnighttime with measured radiometric uncertainties of 3.5%,7.8%, and 11.0%, respectively. The data are calibrated andspectrally normalized with respect to an on-board solar diffuser,which has a reflectance that is monitored for stability. Noise-equivalent radiance increases from ∼1 × 10−10 W·cm−2·sr−1 atnadir to ∼3 × 10−10 W·cm−2·sr−1 at scan edge.The current findings came about unexpectedly during routine

instrument check-out procedures for the VIIRS/DNB. To char-acterize and remove sensor noise and offset patterns, we exam-ined astronomically dark scenes over the open oceans duringnights around the new moon (i.e., completely devoid of sunlight,moonlight, and anthropogenic light contamination). Upon firstinspection of these data, the noise pattern was found to containirregularly distributed structures. The anomalous structures werediscovered to be meteorological clouds (Fig. 1) illuminated by anunanticipated source of visible light.

Sources of Night Sky Brightness and Relative MagnitudesA complete description of the nonlunar nighttime illuminationsources can be found in the reviews of refs. 1 and 3. We shall limitour discussion here to the globally ubiquitous sources (namelyairglow, starlight, and zodiacal light), referring to these collec-tively as “diffuse illumination, multisource” (or “DIM”) emissionsof nighttime visible light.The zodiacal light (11) arises from sunlight scattered by in-

terplanetary dust in the solar system. Because this dust is concen-trated along the ecliptic and increases in density with proximity tothe Sun, there is a strong spatial structure to this source (ref. 3,figure 37). At high elongation angles from the Sun (i.e., near localmidnight, approximately the time of the Suomi NPP overpass), this

Author contributions: S.D.M. designed research; S.D.M. performed research; S.D.M.,S.P.M., and C.D.E. contributed new reagents/analytic tools; S.D.M., S.P.M., C.D.E., D.T.L.,T.F.L., and J.D.H. analyzed data; and S.D.M., S.P.M., and T.F.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. M.M. is a guest editor invited by theEditorial Board.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. E-mail: steven.miller@colostate.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1207034109/-/DCSupplemental.

15706–15711 | PNAS | September 25, 2012 | vol. 109 | no. 39 www.pnas.org/cgi/doi/10.1073/pnas.1207034109

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 28

, 202

0

light source is minimized, with an expected 640-nm radiance rangespanning roughly 0.4–1.1 × 10−10 W·cm−2·sr−1·μm−1 (from eclipticpoles to equator, respectively).Starlight is composed mainly of contributions from the ∼8,500

visible stars and the integrated light from the ∼1011 subvisible starsof the Milky Way. The brightness of the galactic source is highlyvariable (12), with the largest values along the galactic equator andsignificantly smaller values at the poles (ref. 3, table 36). Conse-quently the magnitude of this source varies over the course of theyear, smallest in December and peaking in July. The combinedstarlight source produces 640-nm radiances in the range of roughly1.8–4.0 × 10−10 W·cm−2·sr−1·μm−1. Contributions from the diffusescatter of galactic and cosmic light by interstellar dust are twoorders of magnitude lower and considered here to be negligible.Among the DIM sources, airglow dominates in terms of its

magnitude, dynamic range, and space/time variability, and soreceives special attention here. Airglow refers to the self-illumi-nation of the upper atmosphere via chemiluminescence processes.Nocturnal airglow (called “night glow”) results from the photo-ionization of atmospheric gases by ultraviolet (UV) sunlight. Nightglow brightness ranges from 10−10 to 10−9 W·cm−2·sr−1·μm−1 (13)within theDNB sensor responsewhen viewed from the surface andat local zenith. Whereas the OLS has not demonstrated the abilityto produce useful meteorological imagery from the DIM sources,the slight red shift of the DNB sensor response (toward the sig-nificant near-infrared Meinel OH* airglow bands; see Fig. S1)makes it inherently more sensitive to these emissions.At the surface, the airglow layer appears brightest (a factor of

∼4 larger than the zenith value) at an elevation angle of ap-proximately 10° (the van Rhijn function, e.g., ref. 14). Airglow asseen from space has been documented from the Space Shuttle(15) and more recently from the International Space Station(ISS; http://eol.jsc.nasa.gov/Videos/CrewEarthObservationsVideos).Fig. 2, taken by astronauts on board the ISS, shows the relativebrightness and distribution of several nocturnal light sourcesincluding the airglow layer.

Airglow Spectral/Space/Time CharacteristicsAirglow emissions occur both as distinct bands and as continuumemission, spanning the UV through the visible and into the nearinfrared (16, 17). Significant contributions with respect to the

DNB response include atomic oxygen at 557.7, 630, and 636 nm(altitudes between ∼250–300 km), atomic sodium at 589.0 and589.6 nm, excited hydroxyl (OH*) radicals (500 nm out to 4.5μm), and molecular oxygen from 761.9 nm (A-band) and 864.5nm. The dominant OH* emission layer is geometrically thin (10–20 km), occurring around a critical air density at approximately85 km where excited species are abundant and the favorablemechanism for energy dissipation is photon emission.Airglow emissions are highly variable spatially, seasonally, and

diurnally (18–20) and track changes in solar insolation, atmo-spheric density, and atomic oxygen availability. This behaviorincludes a semiannual oscillation in maxima with amplitudevariations that are lower in the tropics and higher at midlatitudes(20–22). Regionally, the emissions produce complex transientbanding or patch-like structures associated with planetary waves(23) and mesospheric tides (24, 25). At finer spatial scales,emission features are associated with ionospheric disturbancesforced by atmospheric gravity waves (26, 27). The observedstructures have been tied to tropospheric convection and seismicactivity, with the latter proposed as a means for early detectionof tsunamis (28).

Previous MeasurementsThe Orbiting Geophysical Observatory (OGO-4) provided the firstglobal maps of the time-varying airglow distribution and intensity(19). More recent airglow monitoring sensors include the WindImaging Interferometer (WINDII; ref. 24), the Special Sensor UVSpectrographic Imager (SSUSI; ref. 29), the Optical SpectrographIR imager (OSIRIS) limb-viewing camera (30), and the Thermo-sphere Ionosphere Mesosphere Energetics and Dynamics(TIMED) instrument (25). These sensors were designed to char-acterize upper atmospheric properties that influence high frequencycommunications and near-field sources of light contamination thatimpact astronomy, as opposed to meteorological applications.Early photometer observations from OGO-4 suggested a pos-

sible meteorological utility of airglow (31), although the non-imaging and spatially coarse (∼100 km) nature of thosemeasurements permitted only crude inference of transitionsbetween low and high albedo features (e.g., crossing from oceanto desert). Similarly, the DMSP/SSUSI NPS (a nonimagingphotometer) enlists a secondary measurement at 629.4 nm to

30N30N30N

0 0 0

30S30S30S

12

0E

12

0E

15

0E

15

0E

18

01

80

15

0W

15

0W

12

0W

12

0W

90

W9

0W

12

0E

15

0E

18

0

15

0W

12

0W

90

W

AA

AA

LL

CC

CC

LL

Fig. 1. Low-light imagery from a series of adjacent Suomi NPP VIIRS/DNB nighttime passes over the Pacific Ocean on the night of February 22, 2012. Thecoverage domain spans 20,000 km east-to-west and 12,500 km north-to-south, with geopolitical boundaries drawn in green. The data were collected duringnew moon conditions (no sunlight or moonlight present). In addition to city light emissions (e.g., L), the observations capture clouds (e.g., C) illuminated byreflected airglow, starlight, and zodiacal light. Also apparent are broad, diffuse regions of primary airglow emission (e.g., A).

Miller et al. PNAS | September 25, 2012 | vol. 109 | no. 39 | 15707

ENVIRONMEN

TAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 28

, 202

0

characterize and remove background contributions from Earth’salbedo. To date, the only documented use of nonsolar/lunarsources for cloud imaging comes from surface-based camerasystems (32), based on cloud extinction of down-welling airglowemissions as opposed to cloud reflectance.

Example ImageryFig. 1 shows VIIRS/DNB cloud imagery over the Pacific Oceanfrom several consecutive nighttime passes in conditions com-pletely devoid of both sunlight and moonlight. To avoid edge-of-scan noise effects, the data were cropped at a maximum sensorzenith angle of 60°, resulting in small coverage gaps betweenadjacent passes at lower latitudes. City lights appear as discrete,bright features. Additional examples are published (Figs. S2–S4).

To compare the relative brightness of various illuminationsources, we produced normalized distributions of DNB radiance(in W·cm−2·sr−1) from the north/central Pacific Ocean duringnighttime newmoon (i.e., only DIM sources), nighttime full moon,and daytime overpasses. Fig. 3 shows that nighttime scenes illu-minated by full moon are roughly 1 million times fainter than thedaytime, and the new moon scenes are approximately 100 timesfainter than full moon. Secondary modes in the full moon anddaytime distributions correspond to lunar and solar glint (specularreflection off the ocean surface) regions, respectively.Nighttime detection of low clouds and other near-surface fea-

tures in the thermal infrared is problematic because of poorthermal contrast. Here, DIM light imagery offers distinct advan-tages. Fig. 4 compares VIIRS DNB and M15 (10.7 μm) observa-tions of a low cloud layer over the northern Korean Peninsula. Thewestern and northern edges of the low cloud field are difficult todiscern over land in the infrared imagery but stand out readily inthe DIM light imagery. The imagery also reveals details of the lowcloud structure below optically thin cirrus.Fig. 5 shows a DIM light image of a convective system over the

Pacific Ocean. The lightning flash (which appears as a brightsegment oriented along a DNB scan line, due to the scanningnature of the sensor) and convective clouds are features thatcould have been detected with heritage DMSP/OLS sensor.However, in addition to these features, the DNB reveals a trainof mesospheric gravity waves in the primary airglow emission,originating from the area of convection and propagating away tothe east/northeast.

Source Determination and AnalysisWe considered the possibility that thermal emission might explainthe cloud detections. Ground tests of the silicon-based DNBdetectors showed no sensitivity to wavelengths > 0.91 μm, andeven at this limit the levels of blackbody emission are vanishinglysmall—at least 10 orders of magnitude below minimum detectionlimits. Before the thermal infrared VIIRS bands are cooled to their80 K operational temperature, any cross talk with the DNB wouldhave manifested as white noise as opposed to the coherent struc-tures observed, and the character of DNB imagery did not changeas these bands were cooled. Further evidence against a thermal

Airglow

Moon

Aurora

Starlight

City Lights

ISS

Solar

Panels

Fig. 2. Earth’s limb as seen from the International Space Station on February 1, 2012, 06:11:39 UTC, over the northern Atlantic Ocean. The view is to thenortheast, and shows qualitatively the relative brightness and distribution of various sources of nighttime illumination including airglow, the aurora borealis,the gibbous moon, starlight, and city lights. The lunar source is significantly brighter than the DIM sources, appearing like a star in this exposure. A solar panelfrom the ISS appears in the foreground. (Nikon D3s image ISS030-E-73400 courtesy of the Image Science and Analysis Laboratory, NASA Johnson Space Center).

Fig. 3. Distributions of Suomi NPP VIIRS DNB-measured radiances over thenorth/central Pacific Ocean for nighttime newmoon, full moon, and daytimescenes. The secondary low-radiance modes (G) of the full moon and daytimedistributions are due to moon and sun glint (respectively) off the oceansurface—a characteristic not seen in the extended diffuse light sources.

15708 | www.pnas.org/cgi/doi/10.1073/pnas.1207034109 Miller et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 28

, 202

0

leak came from the simple observation that clouds in the tropicsappear brighter than the surrounding clear-sky ocean background.This behavior is opposite to the emissive signature expected fromcool clouds observed against a warmer underlying surface butconsistent with visible light reflectance of clouds observed againsta low albedo ocean background.The imagery in Figs. 1, 4 and 5 appear similar to daytime obser-

vations, which reveals synoptic-scale patterns of migratory stormclouds, maritime boundary layer clouds, and tropical convection.One notable difference from daytime imagery, however, is thepresence of broad (∼1,000 km scale) regions of diffuse brightness.These features, as noted in Fig. 1, are thought to be areas whereprimary airglow emissions are locally strong enough for the DNB todetect directly. Similar structures were observed on all nights but in

different locations. Their scale and distribution are consistent withOGO-4 airglow observations (19).Other characteristics of the DNB imagery speak to the unique

nature of the DIM illumination. Note in Fig. 3 the smooth low-end tail structure that is unique to the new moon distribution.The secondary low-radiance modes seen in the daytime and fullmoon cases come primarily from the solar and lunar glintregions. The lack of such a glint mode in the new moon distri-bution is consistent with the extended diffuse nature of the DIMemissions. The pan-horizon illumination from airglow representsa fundamental difference from the highly directional solar/lunarsources in the sense that cloud shadows will be minimized—potentially improving cloud masks.Concerning the airglow waves observed in Fig. 5, the measured

horizontal wavelength of ∼33 km is consistent with thunder-storm-induced waves observed by the Midcourse Space Experi-ment (MSX) midwave-infrared (4.3 μm) limb sounder data (33,34). Sensitivity to both primary and reflected DIM light sourcesprovides a unique perspective on coupling between the tropo-sphere and mesosphere. Additional examples of these wavescompared against thermal infrared imagery are published (Figs.S5–S7). In all cases, the waves were not present in correspondingVIIRS thermal infrared imagery.

Research and Operational ImplicationsThe information content of most measurements from meteoro-logical satellites falls off markedly at night, particularly with regardto lower tropospheric clouds (owing to poor thermal contrast),which play a key role in defining Earth’s energy budget (35).Sensing based on DIM light sources could improve the nocturnallow cloud climatology. In turn, it could improve the fidelity of thesea surface temperature (SST) climate data record that is funda-mental to our assessments of climate change, because SSTs arederived from nighttime observations and require a cloud screeningthat is inherently problematic for low clouds (36). As shown in Fig.4, the DIM light measurements hold potential to improve noc-turnal low cloud masks and, thereby, improve the uncertaintystatistics associated with this key climate parameter.Low clouds and fog also pose significant hazards to trans-

portation by air, land, and sea. At high latitudes, detecting cloudsover cold land surfaces and identifying snow and sea iceboundaries are particularly challenging tasks, especially duringthe winter months when sunlight is unavailable for extendedperiods. Although multispectral techniques (37–39) in manycases overcome thermal contrast issues in nocturnal low clouddetection, these algorithms face insurmountable difficulties in

Low CloudsLow CloudsLow Clouds

35N35N35N

40N40N40N

12

5E

12

5E

12

5E

13

0E

13

0E

13

0EA B

Low CloudsLow CloudsLow Clouds

35N35N35N

40N40N40N

12

5E

12

5E

12

5E

13

0E

13

0E

13

0E

ThinThin

CirrusCirrus

Thin

Cirrus

Fig. 4. VIIRS/DNB low-light imagery (A) of the Korean Peninsula on the night of February 23, 2012, at 1730 UTC during new moon, showing low clouds on theeastern coast and superior edge contrast over land compared with VIIRS/M15 (10.763 μm) thermal infrared imagery (B). Note in particular the western andnorthern edges of the cloud formation.

15N15N

10N10N

5N5N

15

0W

15

0W

14

0W

14

0W

14

5W

14

5W

LightningLightning

FlashFlash

15N

10N

5N

15

5W

15

5W

15

5W

15

0W

14

0W

14

5W

Lightning

Flash

Fig. 5. DNB imagery of thunderstorms (lightning flash indicated) in thetropical central Pacific Ocean on February 22, 2012, at 1107 UTC (a zoomed-in subset of Fig. 2). Airglow waves are evident in the vicinity of the storms.The waves, which have wavelength λ ∼ 33 km, appear to propagate radiallyoutward from the area of convection and extend far beyond the anvil cirrus.

Miller et al. PNAS | September 25, 2012 | vol. 109 | no. 39 | 15709

ENVIRONMEN

TAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 28

, 202

0

the presence of overriding cirrus. DIM light imagery, with itsability to peer through optically thin cirrus layers, holds uniquevalue for overcoming this problem.Although the VIIRS/DNB has demonstrated the potential of

DIM light, future sensors could be optimized to better exploitthese signals. For example, geostationary-based sensors couldprovide the first continuous (24-h) visible light observing capa-bility. Such time-resolved imagery would allow for monitoringprimary airglow emissions, including mesospheric wave detection(Fig. 5) in support of tsunami warning systems. Low earth-orbiting satellites such as Suomi NPP must conduct short in-tegration times per sample because of their motion relative tothe surface, but with geostationary satellites, the observed sceneis essentially fixed, allowing for much longer integration times.This staring capability would help to overcome the weaker signaldue to significantly higher geostationary orbital altitudes (35,786km), which enables imagery with spatial resolution comparableto NPP while providing more frequent updates.

ConclusionTheeyes of SuomiNPPhaveopened our own to visible-light sourcesthat transcend the darkness and understood limitations of nighttimeenvironmental sensing. The capability of the VIIRS/DNB to detectairglow and starlight illumination holds important and immediatepractical implications for climate assessment, weather and hazardsmonitoring, and our ability to observe interactions between thelower and upper atmosphere. The particular value of airglow, tra-ditionally regarded within the astronomy community as a nuisance,seems to reaffirm the old adage that “One man’s trash is anotherman’s treasure.” These findings stand to influence the scope anddesign of next-generation environmental satellite missions.The complex space/time variability of DIM light sources pres-

ents unique research challenges for quantitative applications.Whereas contributions from the integrated starlight and zodiacallight will vary with both local observation time and season in pre-dictable ways, contributions from the airglow include the primaryemission. A multisensor approach that incorporates simultaneous,independent observations of the primary airglow emission layermay offer a way to quantify this highly dynamic source.

MethodsTo calibrate and quality-control the VIIRS/DNB data, departures from labora-torymeasurements of the background electronic noise and offset patternmustbe characterized and removed (40, 41). The offset for the high gain stage ofthe DNB was determined by observing the Pacific Ocean during the newmoon, when the solar zenith angle was at least 105° (15° below the horizon).Only data between 50° south and 50° north latitude were included to excludeillumination from the aurora. A mask was used to exclude areas of knownanthropogenic illumination. Outliers were rejected by using a 3-δ filter.

Nighttime imagery examples were produced from Suomi NPP VIIRS/DNBdescending nodes by using data obtained from the NPP Integrated DataProcessing Segment. Only newmoon data under astronomical dark (solar andlunar zenith angles > 108°) illumination conditions were considered. Thenoise/offset corrected radiance data were remapped to a Mercator pro-jection, cropped to a maximum satellite zenith angle of 60° to avoid residualedge-of-scan noise patterns, scaled logarithmically between [−11.5, −9.0]log(W·cm−2·sr−1), and plotted by using a linear grayscale color palette.

Normalized distributions of DNB radiance for new moon, full moon, anddaytime scenes were based on data collections over the north/central PacificOcean (no land surface reflectance or city light contributions). New moondata were from February 23, 2012 [1404–1422 Coordinated Universal Time(UTC)], full moon from December 7, 2011 (1220–1232 UTC), and daytimefrom March 10, 2012 (2331–2351 UTC). Total samples in each distributionexceeded 17 million. Noise/offset corrections for the new moon and fullmoon data, and a factor of two bit-stripping correction for the daytimeradiances, were applied. The radiance data for all three cases were cropped ata maximum sensor zenith angle of 60° and logarithmically scaled, binned overthe range [−13.0, 0.0] log(W·cm−2·sr−1) at 0.05 increments, and normalized.

ACKNOWLEDGMENTS. We thank Kohji Tsumura (Institute of Space andAstronautical Science, Japan Aerospace Exploration Agency), ChristophLeinert (Max Planck Institute for Astronomy), and Joachim Köppen (Univer-sity of Strasbourg) for insight on nighttime light sources; Jody Russell [ImageScience and Analysis Laboratory, National Aeronautics and Space Adminis-tration (NASA)-Johnson Space Center) and Dr. Donald Pettit [NASA, Interna-tional Space Station (ISS) Astronaut] for assistance with ISS photography;and Jeffrey Cox (Aerospace, Offutt Air Force Base) for assistance with De-fense Meteorological Satellite Program datasets. We acknowledge the sup-port of the Naval Research Laboratory through contract N00173-10-C-2003,the Oceanographer of the Navy through the Program Executive Office C4l/PMW-120 under program element PE-0603207N, and the National Oceanicand Atmospheric Administration Joint Polar Satellite System Cal/Val andAlgorithm Program. The views, opinions, and findings in this report arethose of the authors, and should not be construed as an official NOAAand/or US Government position, policy, or decision.

1. InghamMF (1971) The light of the night sky and the interplanetary medium. Rep ProgPhys 34:875–912.

2. Roach FE, Gordon JL (1973) The Light of the Night Sky (D. Reidel, Dordrecht, TheNetherlands).

3. Leinert Ch, et al. (1997) The 1997 reference of diffuse night sky brightness. AstronAstrophys Suppl Ser 127:1–99.

4. Croft TA (1978) Night-time images of the earth from space. Sci Am 239:68–79.5. Croft TA (1973) Burning waste gas in oil fields. Nature 245:375–376.6. Welch R (1980) Monitoring urban population and energy utilization patterns from

satellite data. Remote Sens Environ 9:1–9.7. Elvidge CD, Baugh KE, Kihn EA, Kroehl HW, Davis ER (1997) Mapping of city lights

using DMSP Operational Linescan System data. Photogramm Eng Remote Sensing 63:727–734.

8. Miller SD, Haddock SHD, Elvidge CD, Lee TF (2005) Detection of a bioluminescentmilky sea from space. Proc Natl Acad Sci USA 102:14181–14184.

9. Salomonson VV, Barnes WL, Maymon PW, Montgomery HE, Ostrow H (1989) MODIS:Advanced facility instrument for studies of the Earth as a system. IEEE Trans GeosciRem Sens 27:145–153.

10. Lee TF, et al. (2006) The NPOESS/VIIRS day/night visible sensor. Bull Am Meteorol Soc87:191–199.

11. Tsumura K, et al. (2010) Observations of the near-infrared spectrum of the zodiacallight with CIBER. Astrophys J 719:394–402.

12. Toller GN (1990) Galactic and extragalactic background radiation, optical ob-servations of galactic and extragalactic light: Implications for galactic structure. Pro-ceedings IAU Symposium 139, eds Bowyer S, Leinert Ch (Kluwer, Dordrecht, TheNetherlands), pp. 21–34.

13. Hoffman W, Lemke D, Thum C (1997) Balloon-borne infrared telescope for absolutesurface photometry of the night sky. Appl Opt 16:3125–3130.

14. Kwon SM, Hong SS, Weinberg JL (1991) Origin and Evolution of Interplanetary Dust,Proceedings IAU Symposium 126, eds Levasseur-Regourd AC, Hasegawa H (Kluwer,Dordrecht, The Netherlands), pp. 179–182.

15. Mende SB, Banks PM, Nobles R, Garriott OK, Hoffman J (1983) Photographic ob-servations of Earth’s airglow from space. Geophys Res Lett 10:1108–1111.

16. Broadfoot AL, Kendall KR (1968) The Airglow Spectrum, 3100-10,000Å. J Geophys Res

Space Physics 73:426–428.17. Meinel AB (1950) OH emission bands in the spectrum of the night sky. Astrophys J

111:555.18. Donahue TM, Guenther B, Thomas R (1973) Distribution of atomic oxygen in the

upper atmosphere deduced from Ogo 6 airglow observations. J Geophys Res 78:

6662–6689.19. Reed EI, Fowler WB, Blamont JE (1973) An atlas of low-latitude 6300-Å [O I] night

airglow from Ogo 4 Observations. J Geophys Res 78:5658–5675.20. Xu J, et al. (2010) Strong longitudinal variations in the OH nightglow. Geophys Res

Lett 37:L21801.21. Cogger LL, Elphistone RD, Murphree JS (1981) Temporal and latitudinal 5577 Å air-

glow variations. Can J Phys 59:1296–1307.22. Takahashi H, Onohara A, Shiokawa K, Vargas F, Gobbi D (2011) Atmospheric wave

induced O2 and OH airglow intensity variations: effect of vertical wavelength and

damping. Ann Geophys 29:631–637.23. Adachi T, et al. (2010) Midnight latitude-altitude distribution of 630 nm airglow in the

Asian sector measured with FORMOSAT-2/ISUAL. J Geophys Res 115:A09315.24. Shepherd GG, McLandress C, Solheim BH (1995) Tidal influence on O(1S) airglow

emission rate distributions at the geographic equator as observed by WINDII. Geo-

phys Res Lett 22:94GL03052.25. Marsh DR, Smith AK, Mlynczak MG, Russell JM, III (2006) SABER observations of the

OH Meinel airglow variability near the mesopause. J Geophys Res 111:A10S05.26. Hersé M (1984) Waves in the OH emissive layer. Science. New Series 225:172–174.27. Mende SB, Swenson GR, Geller SP, Spear KA (1994) Topside observation of gravity

waves. Geophys Res Lett 21:2283–2286.28. Makela JJ, et al. (2011) Imaging and modelling the ionospheric airglow response over

Hawaii to the tsunami generated by the Tohoku earthquake of 11 March 2011. Ge-

ophys Res Lett 38:L00G02.29. Paxton LJ, et al. (1992) SSUSI: An horizon-to-horizon and limb viewing spectrographic

imager – UV remote sensing, SPIE International Symposium on Optical Applied Sci-

ence and Engineering, Ultraviolet Technology IV, SPIE Paper 1764-15.

15710 | www.pnas.org/cgi/doi/10.1073/pnas.1207034109 Miller et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 28

, 202

0

30. Degenstein DA, Llewellyn EJ, Lloyd ND (2003) Volume emission rate tomographyfrom a satellite platform. Appl Opt 42:1441–1450.

31. Warnecke G, et al. (1969) Meteorological results from multi-spectral photometry inairglow bands by the OGO-4 satellite. J Atmos Sci 26:1329–1339.

32. Walker DE, Schwarz HE, Bustos E (2006) Monitoring the night sky with the CerroTololo all-sky camera for the TMT and LSST projects. Proc SPIE 6267:62672O.

33. Taylor MJ, HapgoodMA (1988) Identification of a thunderstorm as a source of short periodgravity waves in the upper atmospheric nightglow emissions. Planet Space Sci 36:975–985.

34. Dewan EM, et al. (1998) MSX satellite observations of thunderstorm-generatedgravity waves in mid-wave infrared images of the upper stratosphere. Geophys ResLett 25:939–942.

35. Stephens GL (2005) Cloud feedbacks in the climate system: A critical review. J Clim 18:237–273.

36. Patrenko B, Ignatov A, Kihai Y, Heidinger A (2010) Clear-sky mask for the advancedclear-sky processor for oceans. J Atmos Ocean Technol 27:1609–1623.

37. Lee TF (2000) Nighttime observation of sheared tropical cyclones using GOES 3.9-μmdata. Weather Forecast 15:759–766.

38. Ellrod GP (1995) Advances in the detection and analysis of fog at night using GOESmultispectral infrared imagery. Weather Forecast 10:606–619.

39. Pavolonis MJ (2010) Advances in extracting cloud composition information fromspaceborne infrared radiances—a robust alternative to brightness temperatures. PartI: Theory. J Appl Meteorol Climatol 49:1992–2012.

40. Mills S, et al. (2010) Calibration of the VIIRS Day/Night Band (DNB). American Mete-orological Society 6th Annual Symposium on Future National Operational Environ-mental Satellite Systems-NPOESS and GOES-R, 9.4 (American Meteorological Society,Boston). CD-ROM..

41. Jacobson E, et al. (2010) Operation and characterization of the Day/Night Band (DNB)for the NPP Visible/Infrared Imager Radiometer Suite (VIIRS). AmericanMeteorologicalSociety 6th Annual Symposium on Future National Operational Environmental Satel-lite Systems-NPOESS and GOES-R (American Meteorological Society, Boston), P349.

Miller et al. PNAS | September 25, 2012 | vol. 109 | no. 39 | 15711

ENVIRONMEN

TAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 28

, 202

0